Ana-Maria Fuertes

Unobserved heterogeneity in panel time series models

Recently, the large T panel literature has emphasized unobserved, time-varying heterogeneity that may stem from omitted common variables or global shocks that affect each individual unit differently. These latent common factors induce cross-section dependence and may lead to inconsistent regression coefficient estimates if they are correlated with the explanatory variables. Moreover, if the process underlying these factors is nonstationary, the individual regressions will be spurious but pooling or averaging across individual estimates still permits consistent estimation of a long-run coefficient. The need to tackle both error cross-section dependence and persistent autocorrelation is motivated by the evidence of their pervasiveness found in three well-known, international finance and macroeconomic examples. A range of estimators is surveyed and their finite-sample properties are examined by means of Monte Carlo experiments. These reveal that a mean group version of the common-correlated-effects estimator stands out as the most robust since it is the preferred choice in rather general (non) stationary settings where regressors and errors share common factors and their factor loadings are possibly dependent. Other approaches which perform reasonably well include the two-way fixed effects, demeaned mean group and between estimators but they are less efficient than the common-correlated-effects estimator.

New panel unit root tests of PPP

Abstract We use the new panel unit root tests [Im, K.S., Pesaran, M.H., Shin, Y., 1995. Testing for unit roots in heterogenous panels. WP 9526, DAE, University of Cambridge. Revised March 1997] to analyse the stationarity of real exchange rates for the G10 economies and Switzerland. Our results for monthly data 1973–96 indicate mean reversion in real exchange rates and a half life of under three years for one-off shocks.

Early warning systems for sovereign debt crises: The role of heterogeneity

Sovereign default models that differ in their treatment of unobservable country, regional and time heterogeneities are systematically compared. The analysis is based on annual data over the 1983-2002 period for 96 developing economies. Inference-based criteria and parameter plausibility overwhelmingly favour more complex models that allow the link between the probability response and the fundamentals to vary over time and across countries. However, out-of-sample forecast evaluation using several loss functions and equal-predictive-ability tests suggests that simplicity beats complexity. Parsimonious pooled logit models produce the most accurate sovereign default forecasts and outperform the naive benchmarks.

On sovereign credit migration: A study of alternative estimators and rating dynamics

Different estimators of rating transition matrices have been proposed in the literature but their behaviour has been studied mainly in the context of corporate ratings. The finite-sample bias and variability of three sovereign credit migration estimators is investigated through bootstrap simulations. These are a discrete multinomial estimator and two continuous-time hazard rate methods, one of which neglects time heterogeneity in the rating process whereas the other accounts for it. Panel logit models and spectral analysis are utilized to study the properties of the rating process. The sample consists of Moodys ratings 1981-2004 for 72 industrialized and emerging economies. Hazard rate estimators yield more accurate default probabilities. The time homogeneity assumption leads to underestimating the default probability and greater migration risk is inferred upon relaxing it. There is evidence of duration dependence and downgrade momentum effects in the rating process. These findings have important implications for economic and regulatory capital allocation and for the pricing of credit sensitive instruments.

Nonparametric cointegration analysis of real exchange rates

This study indirectly addresses the issue of potential nonlinearities in real exchange rate adjustment for 18 OECD economies 1973–1998 using recent developments in the theory of nonparametric cointegration. While the standard Johansen tests yield mixed evidence, the results from a new nonparametric approach are clearly supportive of real exchange rate stationarity. Since the latter approach allows for a relatively general data-generating process, the findings are consistent with nonlinear mean reversion.

Numerical issues in threshold autoregressive modeling of time series

This paper analyses the contribution of various numerical approaches to making the estimation of threshold autoregressive time series more efficient. It relies on the computational advantages of QR factorizations and proposes Givens transformations to update these factors for sequential LS problems. By showing that the residual sum of squares is a continuous rational function over threshold intervals it develops a new fitting method based on rational interpolation and the standard necessary optimality condition. Taking as benchmark a simple grid search, the paper illustrates via Monte Carlo simulations the efficiency gains of the proposed tools.

Asymmetric dynamics in UK real interest rates

This paper explores the long run behaviour and short run dynamics of quarterly UK real interest rates, 1950–1999, in a threshold autoregressive framework. Using bootstrap LR extensions of the Enders and Granger (1998) threshold unit root and asymmetry tests, it finds support for sign and amplitude asymmetric mean reversion. These findings provide one explanation for the apparent persistence in real interest rates and are consistent with asymmetric feedback rules for inflation targeting.

How do UK Banks React to Changing Central Bank Rates

This paper explores the interest rate transmission mechanism using a broad disaggregated sample of UK deposit and credit products. For a large proportion of rates the adjustment speed is time-varying, switching among four regimes according to the direction of the policy rate and its effect on the disequilibrium gap. In general, this sign asymmetry implies faster adjustment to the long run path when the policy rate revision widens the gap. There is evidence of curvature in the catch-up effect towards equilibrium, namely, large gaps entail a disproportionately faster correction although mainly for deposits. The size of the policy rate change also impacts the adjustment speed. The notable heterogeneity found across financial institutions/products regarding the presence of these nonlinear patterns raises important questions on how to assess the effectiveness of monetary policy. The cross-section heterogeneity uncovered can be explained to some extent by diversification, profit volatility, product range, market concentration and menu costs.

A Non-linear Analysis of Excess Foreign Exchange Returns

In this paper we explore the dynamics of US dollar excess foreign exchange returns for the G10 currencies and the Swiss franc, 1976-97. The non-linear framework adopted is justified by the results of linearity tests and a parametric bootstrap likelihood ratio statistic which indicate threshold effects or differential adjustment to small and large excess returns. Impulse response analysis suggests that the effect of small shocks to excess returns inside the no-arbitrage band exhibits most persistence. Large shocks outside the band decay most rapidly and also exhibit overshooting. These phenomena are explained in terms of noise trading strategies and transaction costs. Copyright 2001 by Blackwell Publishers Ltd and The Victoria University of Manchester

Small Sample Properties of Panel Time-Series Estimators with I(1) Errors

Monte Carlo simulations are used to explore the small-sample properties of a mean group and two pooled panel estimators of a regression coefficient when the regressor is I(1). We compare and contrast the effect of I(0) and I(1) errors and homogeneous and heterogeneous coefficients in a design based on two typical PPP panels. The results confirm that the asymptotic theory is relevant to practical applications. With I(0) errors and homogeneous coefficients, the estimators are unbiased, dispersion depends on the signal-noise ratio and falls at rate T(rootN) as expected. With I(1) errors and no cointegration, dispersion falls at rate rootN. When heterogeneity is introduced with I(0) errors, the dispersion of the pooled estimators falls at rate root N, but that of the mean group continues to fall at rate T(rootN). Finally, the pooled estimators are likely to lead to distorted inference both in the case of I(1) errors and the case of I(0) errors with heterogeneous coefficients case. The mean group estimators are, however, are generally correctly sized.